Know any good legibility tests for typefaces?

[Nick Shinn]

I can in fact predict that a page of text will be unreadable if the size of type is such that the counters are too small for my aging eyes to resolve them into distinct shapes.

That's legibility, not readability.
These words have well-established meanings for typographers.
But even for legibility, factors such as lighting and paper stock play a role.

Experimenters might be able to confirm where my threshold is relative to a particular font.

Whether you find a document more or less readable has little to do with whether the particular typeface is set 9 or 10 pt., but on the overall typography of the page. One typographer will set 9 on 14 rag right two columns, whereas another will set 10 on 12 justified three columns. What has size to do with the relative readability of these two pages?!

you are applying standards of art to science,

I'm talking about science and design.
You're trying to discredit my argument against bad science by saying that it's too artistic.

The main thing is that you can measure and test empirically whatever you assert.

And you can't, if you're attempting to isolate the effect of things like serif size or weight on readability.
These things can't be measured.
Does Adobe Garamond have bigger serifs than Times Roman? That's not even a question, and yet it represents the kind of thing that, in theory, you are assuming is relevant to readability.

How can the size of typeface counters be measured?

[Chris Dean]

Track

[ebensorkin]

the science of the fall-down-ability of falling objects

You killed me with this. I was giggling for several minutes. Thanks!

if you’re attempting to isolate the effect of things like serif size or weight on readability.

This is exactly right. But it isn't the case that we have to be gunning for "on readability" - we could go for specific measures like impact on speed for instance and get something which may (or may not) lead to the deeper understanding Peter is after, but which may be interesting all the same.

So instead you have:

if you’re attempting to isolate the effect of things like serif size or weight on speed...

The obvious next question to ask is what targets might you reasonably go for: Serifs strike me as especially complicated. The white space between letters vs that held in counters might be a great one. It would require an explicit formula though. David Kindesley made a device to do this that wasn't perfect but was decent. Adobe has one they call 'optical' that similarly isn't perfect. But these, or a new one might be used. Descender length is probably low hanging fruit. The key thing would be to use a face & then make modifications to it so as to isolate the variables as much as possible.

I should also point out that plenty of perfectly good research is done which doesn't isolate EVERY variable but which notes the ones that have not been isolated along with the ones that have & how they isolated the ones they think they got. Obviously the more the better but...

[William Berkson]

>How can the size of typeface counters be measured?

By the total area of white within the counter. You do have to decide for a letter C how far to go out, but you can make a choice and see whether it turns out to be informative.

You can also measure area within the black in the serifs, the extension from the stem, etc. Actually David Kindersley already did a version of such measurements.

I don't know what's going to be important, but you are quite wrong about the impossibility of measuring this kind of stuff. That's actually the easy part.

I am not opposed to any argument against bad science. I am just looking at the wonderful expose by Gary Taubes of bad science in the field of obesity prevention. It is a truly horrifying story, and seems quite true.

While I agree with you that there there is a great deal of bad science out there, including in the field of reading, it doesn't justify opposing all science. For good science can be a boon to humanity, including artists.

>saying that it’s too artistic.

No, I'm saying you're mixing up artistic and scientific goals, and wrongly applying standards from one to the other.

[Nick Shinn]

you are quite wrong about the impossibility of measuring this kind of stuff. That’s actually the easy part.

No, YOU are quite wrong!

the total area of white within the counter.

Here are two cuts of Futura, URW and Adobe, with x-height a different proportion of the em square.
What is the area of the enclosed counter of this typeface?

Here is Akzidenz Grotesk Bold and Helvetica Bold.
These would appear to have the same counter area, but that tells one nothing about the readability of the face en masse.
And how would one measure that area--as a ratio of what? Em square?--But definition of the em square is contentious to begin with, as the thread at Typophile last year demonstrated.

[ebensorkin]

Nick, your points would go down so much better at this point if you would stop using this term "readability".

Also I reject the idea that you could do "Bill Style" research on two Typefaces. You would need to have a standard ( control) and a variant or a series of variants. Otherwise you have no basis for looking at a typographic feature.

Of course if you want to just do a speed test of one font vs the other you could do that but as you point out you would have to do them again & again with multiple leadings etc to find their "sweet spot". And basically the data you got out of that would probably not be easy for most people to take in or make use of. If you did a simple X inches wide 10/12 test - easy by comparrision you would be testing in a typographically naive way.

If you wanted to look at opening/appature sizes you would ( I think) want to look at them in the context of interior sizes vs interletter spaces as a sequence of ratios - not as simple interior measurements to see if the Noordzij model for that which seems so "bang-on" bears out. Peter even has a name for this and some theory built up as well. And NO, it isn't going to give you the picture. Nobody would claim that it would.

[William Berkson]

>stop using this term “readability”.

Well I know it is an inexact term, but it's a good start as it's a meaningful word in ordinary language. I don't know why you are so allergic to it. As progress is made on the issue, then more refined terminology will be introduced.

>these would appear to have the same counter area, but that tells one nothing about the readability of the face en masse.

Nick, you are muddling different issues. I said it is easy to measure the counters, and I can tell you how many pixels there are in each in each counter on my monitor, if you like.

Now you are changing the issue and saying that won't tell us about readability en masse. But I already said that measuring a counter is the easy part. It is also easy to measure the ratio of counter to black in the letter, the ratio of black to white in the em square, and so on. The difficult part is identifying which or combination of variables are really important for reading. That's where having good theory, such as Peter is working on, is going to help guide researchers.

You're seem to be saying that you know before research starts that there is no such variable or combination of variables possible to discover that gives us useful information about readability.

It's called research because we don't know where it will lead when we start. Being against starting seems to me a very poor idea.

[ebensorkin]

I don’t know why you are so allergic to it.

Because what you & Nick mean by it are different things. For a term to be useful it has to have a shared meaning. This one doesn't.

[Nick Shinn]

Also I reject the idea that you could do “Bill Style” research on two Typefaces.

Doesn't that mean it's impossible, as I have said all along, to measure the effectiveness of typefaces?

You would need to have a standard ( control) and a variant or a series of variants. Otherwise you have no basis for looking at a typographic feature.

Of course. That's why, as I've argued somewhere above, with regard to serifs, it would make more sense to "control" the typeface. But say you pick Meta, aren't there many, many ways that serifs could be "added" to it, e.g. slab, bracket, etc.--as in the Panose list?

And just suppose you want to measure Speed, Comfort and Comprehension as they relate to counter size, what model are you going to use for varying counter size?

You can't just willy nilly vary counter size and leave other parameters unchanged, because that defeats the whole purpose of having diffent counter sizes. Consider these three Slimbach faces. I've equalized x-height. When the size of the enclosed counter on the e is varied, other proportions in the typeface must also vary for optimum what-ever-it-is-you-want-to-measure. So these must be altered too: the size of the aperture to the lower counter, and the relative size of the capitals and their counters, and even the genre of the type, for a large upper counter on an e is not good for an oldstyle. The size of the e's counter is not a measurement, it's a style.

[Nick Shinn]

You’re seem to be saying that you know before research starts that there is no such variable or combination of variables possible to discover that gives us useful information about readability.

Yes.

[enne_son]

Nick what was behind my claim is that there is a relationship between (1) perceptual discrimination affordance at the granularity of letter parts and (2) rapid, automatic visual word-form resolution affordance at the granularity of word wholes. I relate the first to legibility, and the second to readability.

What I think I get from your descriptions is that, in your way of thinking, legibility has to do restrictively with the perceptual processing factors. And readibility has to do with the total picture, or rather, specifically with all those other ‘look’ dimensions that go beyond these, but aren’t really directly touched by them.

Readibility in this scheme doesn't have it's own distinct perceptual processing component. It is totally a matter of humanistic virtues: aesthetics, optical-grammatical integrity, expression, historical reference, transparency to information rank, contextual propriety.

I say it is these too, but also has a distinct and foundational perceptual processing component. And I’d say this perceptual processing component is sensitive to things like spacing, counter size, relative proportion of ascenders and descenders to x-height. Because it’s a good deal more than that means that gauges of the sensitivity of this percptual processing component to spacing, counter size, etc will always fall too short as reliable indicaters of the total picture.

You go on to try to show that countersize is style and nothing more, but I say it also has a perceptual discrimination affordance component.

[ebensorkin]

Doesn’t that mean it’s impossible, as I have said all along, to measure the effectiveness of typefaces?

I have answered this in detail before. In redux: If you mean as of now; in a scientific way; then yes; we haven't done the research. If you mean in an absolute sense - then yes. Absolute isn't on the menu. If you mean that you couldn't maybe do billions ( or however many) of tests & then maybe come up with something interesting & useful then - no. You can't know that.

But what I meant by this is that focusing on font vs font is not a useful way of thinking about this. That's why I reject it.

That said; it's also true that people make subjective choices about this all the time & will. Thinking "Does this seem more readable to me - or this?" is sometimes a part of font choice and isn't going away.

And in addition they might even make experimentally flawed tests to justify those choices. Wanna stop em? Try it. Good Luck!

It's also true that you can plunk two typefaces into MS Word with default settings ( size+leading ); set a nice long text and then test some factor such as speed with or with comprehension. Will that tell you which face is "better" or "more readable"? No. Of course not. Will it tell you something? Yes! Obviously something about that set of variables. But a very fuzzy something. That is of course not nothing. But it ain't much either. You would have to do many thousands of additional tests to get somewhere that way. That would be expensive and inefficient. And I guess that you would end up with a data set that is too murky to be useful. So again. Impossible No. Very unlikely? Yes.

You can’t just willy nilly vary counter size and leave other parameters unchanged You know that I agree. But you also know, or should by now that this is a yet another straw man at this point.

Unless you are willing to grip specifics in one sense or another you get nowhere. Which is where this thread is now going - fast.

[Scalfin]

Nick, you seem to be saying that because fonts differ in more than single ways. That is where you prove yourself to be an idiot. It is also why god invented the SRS! You take a lot of fonts, controlling for size by taking a large block of text and setting the font sizes as those at which the size of the block of text is constant, and score how easily read the various fonts are, At which point you can arrange the scores for the fonts by various variables. At that point, all you have to worry about is variables associating.

Also, you insist that readability and legibility are different. How about we ask the American Heritage Dictionary:
read·a·ble (rē'də-bəl) pronunciation
adj.

1. Easily read; legible: a readable typeface.

[William Berkson]

Well for what it's worth, here's on the word 'readable'. It originally and still generally means "easily read", as you quote, Scalfin. It could refer to something well written, or it could mean that the printed page is visually easy on the eyes. Among those who have written about typography, readability came to be distinguished from legibility, because these writers thought that there was more to making a text block easy to read than having the letters clear and distinguishable from one another. This specialized meaning evolved by the 1960s, but I wouldn't be surprised if it were much earlier. Whether this is a distinction without a difference is still debatable.

Nick and I agree that there is more to readability than the clarity of individual letters. Where we don't agree is whether this difference is worthwhile investigating scientifically. I agree with Eben that there's not more to be said on it at this point. The research will happen or it won't.

[ebensorkin]

Scott, you are making this faaaar simpler than it is. In pretty much each aspect you have presented.

And you are calling people names. This is worse.

Name calling can eventually be seen as a personal attack. Personal attacks are not acceptable on Typophile.

Also, it doesn't promote healthy dialog. Healthy dialog is what Typophile is for.

Please read this: http://typophile.com/readme

And lastly, IMNSHO the American Heritage Dictionary is better off asking us.

[Scalfin]

Sorry about that, but such testing methods have been used on things more complicated than this.

[enne_son]

Scalfin, you sound like you have some understanding of empirical testing.

How would you propose we operationalize ‘easily read’. And isn't variables associating precisely what is being claimed?

[eliason]

Among those who have written about typography, readability came to be distinguished from legibility, because these writers thought that there was more to making a text block easy to read than having the letters clear and distinguishable from one another. This specialized meaning evolved by the 1960s, but I wouldn’t be surprised if it were much earlier.

For what it's worth, Dwiggins in 1947 does not seem to be making that distinction:

A type face is good if it is easy to read. No concession that interferes with ease of reading may be made either to beauty of appearance or to mechanical felicity. Legibility is the basic law, the sine qua non.

[John Hudson]

Nick, science can help improve our knowledge of anything.

[Scalfin]

There are several ways. We could ask a person to read something then answer some questions about the text and look at how quickly they read. We could look at how often they confuse character, as was done in a test looking at the relative legibility of Constantia, Cambria, and TNR. The only problem is that it would take a large study size to get a satisfyingly low p-value.
By characteristic associating, I mean that fonts w/ open counters usually have long descenders (hypothetical, you could probably name real examples of something like this), thereby confounding any findings on either.

[enne_son]

Scaflin, what part of readibility or legibility would your scores give us information about? Its perceptual processing component? And by that should we mean only what occurs inside the visual cortex? What about the other dimensions of readability that typographers are convinced exist?

Variables associating is exactly Nick’s point. He regards the confounding effect as insurmountable when it comes to isolating factors in any useful way. An disassociating them for the purpose of our tests creates bad type.

[Nick Shinn]

science can help improve our knowledge of anything.

LOL.

[Scalfin]

Because The Passion was really informative, and the quote in the bible of Jesus say "he who is not with me is against me" has really improved diplomacy.

[ebensorkin]

science can help improve our knowledge of anything.

This is an under-dissected phrase. As such; there is no point in a yes/no dichotomy.

[Nick Shinn]

science can help improve our knowledge of anything.

Research and development that fails in its primary goal may yet come up with useful accidents and spin-offs.
Penicillin and stainless steeless steel, for example.
And then there is the drug development strategy obliquely described by the protaganist of Quadrophenia, "I'll do it, what is it?" -- in other words, lets see what effect this chemical has, perhaps it will cure something. Not a bad strategy for type design, BTW, as art directors can come up with quite unexpected uses for fonts.